A review of ten imputation methods for handling missing values in logistic regression

Abstract

This paper presents a brief review of ten imputation methods for missing data in the 
binary logistic regression model. The performance of these methods under different 
missingness scenarios has been examined based on a medical dataset. The results indi
cated that, in general, expectation–maximization and k-nearest neighbor imputation 
methods are very appropriate for estimating the missing values in this model, whether 
data are missing in dependent variable only, independent variables only, or in both.